Multi-objective optimization for antenna design
The problem consisted of the optimization of a seven element antenna array to improve the radiating performance. The three objectives considered are the minimization of the return loss, together with the minimization of the cross polarization and the mutual coupling between the ports; simultaneously, the design should satisfy a complex constraint on the far-field pattern. Since there is no single optimum to be found, the MOGT and MOGA-II (Game Theory-based and Genetic Algorithm) were used as multiobjective algorithms. The optimization of the antenna was carried out by employing a parametric CST MICROWAVE STUDIO (CST MWS) model, performing a time domain analysis, and using ESTECO’s modeFRONTIER4 multi-objective optimization and process integration tool. The distributed optimization search exploited the parallelization capabilities of the MOGT and MOGA-II algorithms, which allowed the simultaneous evaluation of several design configurations by running concurrent threads of the solver. The results obtained are very satisfactory, and the procedure described can be applied to even more complex antenna design problems.
Once the integration of the model in the optimization environment is done, all the optimization capabilities can be applied to improve the performance of the antenna.
The optimization is a full batch process (no human intervention during the run phase), but can be constantly monitored from a “run-log” graphic console. The optimization can always be manually stopped if no further improvements are expected. Today’s multi-core PCs and clusters are able to carry out multi-objectives optimization of a complex model by a reasonable computational time.
Thanks to its clear, well-defined, and non-weighted approach, multi-objective optimization helps in understanding the physics of the problem while exploring the design space, looking for the set of optimal configurations. Using modeFRONTIER4, the antenna engineer who is usually simulating with CST MWS Transient or Frequency domain solvers, can obtain an enhanced benefit: he can now proceed from analysis to efficient optimization and then synthesis